Research On Technology Of Detection And Tracking Of Small And Dim Target In Infrared Image Sequences

Small and dim target detection and tracking in infrared image sequences are extensively applied in many fields such as making to accurate lead, automatic control etc. As the infrared target has far distance from imaging systems with the characteristics that the area is small and the SNR is low etc., it is much more important and difficult to do the job of infrared image preprocessing and detection. How to take full advantage of the infrared image preprocessing and target detection techniques is an important factor in the field of infrared imaging detection. The problem of detecting and tracking small, moving, low amplitude objects in infrared image sequences that also contain moving nuisance objects and background noise is addressed in this thesis after the introduction of current development of small target detection and tracking both domestic and abroad. The main job of the thesis can be described as follows:First, Several image preprocessing methods are introduced. Thresholding segmentation plays an important role in the process of the small target detection. In view of the obvious wrong segmentation in the existing threshold selection method based on two-dimensional histogram vertical segmentation, a new two-dimensional histogram oblique segmentation method and its fast iterative algorithm of the Otsu threshold is proposed in order to improve the performance and efficiency of the original 2D Otsu threshold algorithms.In the part of target detection, using pixel gray information on the one hand, the new dim target detection approaches based on Contourlet transformation and NSCT are put forward. According to the features of the decomposing images after transform, the background clutter and noise are eliminated and the final detecting of the target is realized. On the other hand, in strong undulant clutter background utilizing the characteristics of local statistics background signal and objectives of movement, a method of target detection in infrared image sequences based on the least absolute deviation background predication and chaos-genetic algorithms is proposed as well as a method based on variable neighborhood and step-size LMS algorithm by adaptive prediction of background.In the following tracking process, aiming at difficult problems of small target tracking technology, a tracking algorithm based on Kalman prediction filter and template matching is brought forward to realize the goal of predicting and tracking small target.The effectiveness of the techniques is demonstrated by applying the resulting algorithms to real world infrared (IR) image sequences containing targets of opportunity.